why i left OMGPOP/zynga

christopher william holt bio photo By christopher william holt

My last day at OMGPOP/Zynga was August 2nd, 2012. In order to explain why I left, it’s probably best to understand exactly the function I performed first. When I began work in August 2010 we had roughly 5 million registered users (As I am not sure of the specifics I am not going to mention MAU/DAU metrics). Our sole web property was OMGPOP.com and it looks roughly the same now as it did then. For those 5 million users we had 3 Backend engineers, and I was lucky to be on that team. It consisted of Jason Pearlman, whom I’ve known now for 15 years (nearly half of my life), and Chul Park. For those two years not a day went by where there wasn’t a challenge that we sought to overcome by implementing elegant, albeit usually scrappy, solutions.

Just before the launch of Draw Something (DST) we were somewhere around 22 million registered users, and still the same 3 Backend engineers. Looking back, the time spent at OMGPOP before DST just seems like a bunch of friends cutting their teeth on some fun Internet experiments. When DST launched that all changed, and it became a “scale or die” situation. (The latter was by no means an acceptable option) Most of my memories between the launch and the date of my departure is quite a blur. I think I gave up a fair amount of my personal “shelf space” of memories in order to retain our ever-expanding and changing architecture, which allowed me to rapidly consider new creative solutions for the relentless demand. There were many times we had to rewrite a core service in a new architecture, or even a new language, on the fly and say, “OK, that will scale to at least another 5x,” then go to sleep, wake up, and find that the demand was already at more than 10x. This would repeat. It was an amazing feeling to continually overcome these extremely complex challenges with such minimal resources. Little did we know we actually built an infrastructure that could grow far beyond anyone’s initial expectations of what was needed. There was no single command or magic approach we used to do this, it was only through those long days and hard work that a natural balance between all of our systems came to be.

There have been many numbers discussed regarding DST’s success, but my personal favorite will always have to do with the concurrency that we reached. For many days we achieved a constant rate above 3000 drawings completed per second. On average, each drawing took 30 seconds to complete, which means each second there were 25 hours of content being created. I believe it was also right around this time that Google launched their “One Hour Per Second” marketing campaign for videos being uploaded to YouTube. For me, this put into perspective what we were doing. While I would never directly compare ourselves to Google, as YouTube is an entirely different beast altogether, the numbers spoke loudly.

After weeks of exhausting hours, sometimes getting only 30 minutes of sleep a day, game plays started to display a more predictable growth pattern. (Mostly because people were no longer playing for an excessive 3 hours per day) At this point we were able to relax a bit. At some point during all of this activity the company was bought by Zynga. When I was being told about the news I believe I had my Macbook with me and was not paying attention, as I was probably dealing with some new problem that presented itself earlier that morning. When I began to realize what was being said I slowly shut the lid, and knew that I could safely pause for a few minutes.

That feeling of pause did not last too long. Within a few months after the acquisition we migrated our entire infrastructure over to Zynga’s private data centers. During this migration we documented in detail all the “moving parts” as well as all the different “scale points.” We handed off most of the daily support of basic issues and routine maintenance to Zynga’s extremely talented and personable engineering support teams. At this time things, from a backend development perspective, started to slow down. I now had the time to reflect and fully understand the incredible feat we achieved. I knew I didn’t want the energy to stop. I could hardly sleep, and when I did it was for way too long. My back started to hurt, and for a couple weeks I had trouble walking and doing normal activities. It was like being an addict going through withdrawal. Once I recognized this I knew I couldn’t stay. The one thing I knew I would miss was the culture. If you take what I’ve just written and replace all of the technical nonsense with friendship, and all of the hard work with happiness, that’s exactly what the culture was.

During these two years we were building an amazing team of over 50 people who would do anything for each other. If one person started slow clapping, the entire office would join and eventually erupt into cheers and whistling—we never needed a reason to celebrate. This was before, during, and after DST. DST was a product of the culture that we all built together, and that culture was bulletproof. I think Zynga knew this, and they respected the hell out of us, so we only continued to get stronger once we had some awesome parents in the valley to support us. I knew that these fellow ‘Poppers, who I would get La Colombe coffee with at 9:45am after each of our morning scrum updates, would be lifelong friends. It tore me apart to leave, but while I’m still relatively young I knew I needed to take risks and possibly make mistakes now, rather than later in life. I left what was essentially my family for this newfound addiction of pushing myself to extremes, that had been presented to me from what might have been a singular life opportunity. I really hope it works out, because it’s hard to imagine life being any better than it was, at least as far as I can remember of my times spent in a happy blur.